Spread Prediction and Classification of Asian Giant Hornets Based on GM-Logistic and CSRF Models

Abstract

As an invasive alien species, Asian giant hornets are spreading rapidly and widely in Washington State and have caused significant disturbance to the daily life of residents. Therefore, this paper studies the hornets’ spread and classification models based on the GM-Logistic and CSRF models, which are significant for using limited resources to control pests and protect the ecological environment. Firstly, by combining the improved grey prediction model (GM) with the logistic model, this paper proposes a GM-Logistic model to obtain hornets’ spread rules regarding spatial location distribution and population quantity. The GM-Logistic model has higher accuracy and better fitting effect when only a few non-equally spaced sequences data are used for prediction. Secondly, a cost-sensitive random forest (CSRF) model was proposed to solve the problems of hornets’ classification and priority survey decisions in unbalanced datasets. The hornets’ binary classification model was established through feature extraction, the transformation from an unbalanced dataset to a balanced dataset, and the training dataset. CSRF improves the adaptability and robustness of the original classifier and provides a better classification effect on unbalanced datasets. CSRF outperforms the Random Forest, Classification and Regression Trees, and Support Vector Machines in performance evaluation indexes such as classification accuracy, G-mean, F1-measure, ROC curve, and AUC value. Thirdly, this paper adds human control factors and cycle parameters to the logistic model, obtaining the judgment conditions of report update frequency and pest elimination. Finally, the goodness-of-fit test on each model shows that the models established in this paper are feasible and reasonable.